• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2003년도 학술발표회논문집(1)
• /
• pp.163-170
• /
• 2003

In this study, paying much attention to notable features obtained from spatial distributions of strongly related indices (precipitable water, convergence of air, convective available potential energy) with precipitation, fatal problems in selecting strongly related indices with observed precipitation in a BAIU season were discussed. These results showed spatial distribution of a predicted index provided alternative and physically consistent interpretation for selecting dominant index for heavy rainfall even if the predicted index did not correlate with observed rainfall at a specific observational point as confirmed by the features of CONV (Convergence) or even if it correlated with observed rainfall as confirmed by those of PW (Precipitable Water). Therefore, dominant meteorological indices of heavy rainfall should be selected according to physically evidenced interpretation on features of spatial distributions of indices, and physically and statistically consistent relationship should be built up.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2017년도 제55차 동계학술대회논문집 25권1호
• /
• pp.205-206
• /
• 2017

본 논문에서는 주차 시설 이용 현황 중에서 차량 번호판 조사를 통하여 평균주차시간과 주차 회전율을 구하도록 한다. 관찰하는 사람이 없이 카메라를 활용하고, 조사 시간에 일정한 간격을 주어 조사하도록 한다. 일정한 조사 시간 간격을 주차된 차량에 나눠주어 평균 주차 지속 시간을 구하도록 한다. 그리고 이렇게 하여 주차면 1개당 1시간당 주차 차량 대수인 주차 회전율을 구하도록 한다.

• 대한임베디드공학회논문지
• /
• 제11권4호
• /
• pp.219-225
• /
• 2016

Visual object tracking is one of the key tasks in computer vision. Robust trackers should address challenging issues such as fast motion, deformation, occlusion and so on. In this paper, we therefore propose a visual object tracking method that exploits inter-frame correlations of convolutional feature maps in Convolutional Neural Net (ConvNet). The proposed method predicts the location of a target by considering inter-frame spatial correlation between target location proposals in the present frame and its location in the previous frame. The experimental results show that the proposed algorithm outperforms the state-of-the-art work especially in hard-to-track sequences.

• Journal of Semiconductor Engineering
• /
• 제1권1호
• /
• pp.31-37
• /
• 2020

We present an approximate DRAM architecture for energy-efficient deep learning. Our key premise is that by bounding memory errors to non-critical information, we can significantly reduce DRAM refresh energy without compromising recognition accuracy of deep neural networks. To validate the key premise, we make extensive Monte-Carlo simulations for several well-known convolutional neural networks such as LeNet, ConvNet and AlexNet with the input of MINIST, CIFAR-10, and ImageNet, respectively. We assume that the highest-order 8-bits (in single precision) and 4-bits (in half precision) are protected from retention errors under the proposed architecture and then, randomly inject bit-errors to unprotected bits with various bit-error-rates. Here, recognition accuracies of the above convolutional neural networks are successfully maintained up to the 10^-5-order bit-error-rate. We simulate DRAM energy during inference of the above convolutional neural networks, where the proposed architecture shows the possibility of considerable energy saving up to 10 ~ 37.5% of total DRAM energy.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2016년도 제53차 동계학술대회논문집 24권1호
• /
• pp.105-106
• /
• 2016

본 논문에서는 S/W 개발 보안 지침에서 설계 단계의 보안 코딩 지침을 알려준다. 크로스 사이트 스크립트 공격 취약점(XSS)에서부터 자원 삽입 까지 S/W 보안 취약점의 주요 내용을 입력 데이터의 검증 및 표현에 맞추어 지침을 전달하도록 한다.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2016년도 제53차 동계학술대회논문집 24권1호
• /
• pp.107-108
• /
• 2016

본 논문에서는 승강기 내에서 사람과 사람 사이의 접촉되는 부분을 추출하고자 한다. 승강기에 사람이 많이 타는 경우에는 접촉 현상은 흔히 발생하지 않는다. 하지만 승강기에 사람이 한 두 명 정도 적게 타는 경우에는 종종 발생하곤 한다. 신체적 접촉을 추출하기 위한 방법은 영상을 이진으로 변환하여 이 이진영상에서 뼈대를 추출하여 접촉 여부를 판단한다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2018년도 하계학술대회
• /
• pp.84-87
• /
• 2018

Handwritten letter recognition is important for numerous real-world applications and many topics like human-machine interaction, education, entertainment, and more. This paper describes the implementation of a real-time handwritten letters recognition system on a common embedded computer. Recognition is performed using a customized convolutional neural network, which was designed to work with low computational resources such as the Raspberry Pi platform. The experimental results show that the proposed real-time system achieves an outstanding performance in the accuracy rate and the response time for recognition of twenty-six handwritten letters.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권7호
• /
• pp.2407-2424
• /
• 2022

Recently, Transformer has made great progress in video retrieval tasks due to its high representation capability. For the structure of a Transformer, the cascaded self-attention modules are capable of capturing long-distance feature dependencies. However, the local feature details are likely to have deteriorated. In addition, increasing the depth of the structure is likely to produce learning bias in the learned features. In this paper, an improved Transformer structure named TransDCS (Transformer with Dynamic Convolution and Shortcut) is proposed. A Multi-head Conv-Self-Attention module is introduced to model the local dependencies and improve the efficiency of local features extraction. Meanwhile, the augmented shortcuts module based on a dual identity matrix is applied to enhance the conduction of input features, and mitigate the learning bias. The proposed model is tested on MSRVTT, LSMDC and Activity-Net benchmarks, and it surpasses all previous solutions for the video-text retrieval task. For example, on the LSMDC benchmark, a gain of about 2.3% MdR and 6.1% MnR is obtained over recently proposed multimodal-based methods.

• 방송공학회논문지
• /
• 제27권7호
• /
• pp.1011-1020
• /
• 2022

Recent high-performance panoptic segmentation models are based on transformer architectures. However, transformer-based panoptic segmentation methods are basically slower than convolution-based methods, since the attention mechanism in the transformer requires quadratic complexity w.r.t. image resolution. Also, sine and cosine computation for positional embedding in the transformer also yields a bottleneck for computation time. To address these problems, we adopt three modules to speed up the inference runtime of the transformer-based panoptic segmentation. First, we perform channel-level reduction using depth-wise separable convolution for inputs of the transformer decoder. Second, we replace sine and cosine-based positional encoding with convolution operations, called conv-embedding. We also apply a separable self-attention to the transformer encoder to lower quadratic complexity to linear one for numbers of image pixels. As result, the proposed model achieves 44% faster frame per second than baseline on ADE20K panoptic validation dataset, when we use all three modules.

• 열처리공학회지
• /
• 제35권2호
• /
• pp.82-87
• /
• 2022

The PZT piezoelectric ceramic on the copper alloy plate shows a extraordinary pattern resulted from the electromagnetic and thermal radiation induced by copper coil. The Eddy current or magnetic field by the polarization of PZT grains contained Pb, Zr, Ti with oxide is performed to show the change of array pattern at piezoelectric grains, especially wave-shaped or wrinkled configuration along with lines of electromagnetic field are becoming larger than before while applying the coil induction.